idkmesh

Ownership Concentration: First Real Measurement, and What Structural Debt Refused to Say

Date: 2026-08-30 Evidence level: bounded observational snapshot; no causal claim, no policy authority. Artifacts: results/collaboration/snapshot-2026-08-29T22-03-24Z.json, results/collaboration/observables-2026-08-29T22-03-24Z.json

Why this exists

The first production collaboration-observables snapshot separated two kinds of zero. Three metrics read zero because a real measurement returned nothing. Two read zero because the collector never looked: scripts/collaboration_snapshot.py wrote "changed_file_owners": [] and "structural_debt_finding_ids": [] unconditionally and declared both gaps in its own limitations list.

Collector v0.2 closes both gaps. This report is the first run with them closed.

Window

Fifty most recently created pull requests, #288 through #338, all closed, cutoff 2026-08-29T22:03:24Z. Structural debt was inventoried by tools/idkgraph_observatory.py at revision f9ec91653834a6a318f50434b9e61b69fded0c6c.

Result 1 — ownership concentration is 1.0

Quantity Value
Changed-file observations 281
Attributed to a within-window owner 66
Unattributed (no prior merged touch in the window) 215
Distinct owners 1
Ownership HHI 1.0

Ownership here is a declared model, last_merged_toucher_within_window-v1: a path is owned by the author of the most recent merged pull request inside the window that changed it, and the model is advanced only after a pull request merges, so no pull request is ever credited with ownership it acquired by its own merge.

HHI of 1.0 is the maximum the index can take. Every file whose ownership could be established from the observed window is owned by the same actor. The repository has asserted single-maintainer concentration qualitatively in several places; this is the first time it is a number produced by a collector rather than a sentence produced by a person.

The 215 unattributed observations are not a defect to be tuned away. They are the window boundary made visible: those paths were touched once inside the window and never before it, so the window carries no evidence about who owns them. A collector that guessed an owner for them would be manufacturing the very concentration it is trying to measure.

Result 2 — the deterministic debt is older than the window

Four deterministic findings were loaded, all severity notice, all of one category (document_referenced_only_by_non_markdown_artifact):

Finding Path
debt:01f073c92450dd93 docs/acceptance/PR91_CONTROLLED_DOCKER_GATE.md
debt:e43a860112808597 docs/algorithms/ACO_STIGMERGIC_TASK_ROUTING.md
debt:79ce72d82f955b84 docs/algorithms/HOMEOSTATIC_STIGMERGY_ROUTING.md
debt:8abf4cf4deb90d50 docs/protocols/CONSTITUTIONAL_EVOLUTIONARY_MESH.md

Zero of the four were attributed. No pull request in #288#338 touched any of those four paths. Fifty consecutive pull requests of repository activity passed without one of them going near the deterministic debt that exists.

The trap this run walked into, and the guard that caught it

The observable still reports structural_debt.observed_findings: 0 — byte-identical to the value it reported when nothing was collected at all. The two zeros mean opposite things: the first meant “not looked at”, this one means “looked at, and none of it falls inside the window”.

inventory_complete is what separates them, and it is now derived rather than hard-coded. The collector sets it true only when a report was supplied and every finding in it reached a pull request. Here one condition held and the other did not, so it stays false and structural_debt_findings_outside_the_window_are_unattributed is added to the machine-readable limitations. The snapshot additionally carries collection.structural_debt.findings_loaded: 4 and the full finding index, so the shortfall is recoverable rather than merely flagged.

A reader who takes observed_findings alone will undercount this repository’s deterministic debt by four. That is a real weakness in the observable’s shape, not in the collection, and it is recorded here rather than quietly fixed by inflating the count with findings that have no pull request to attach to.

Limitations

Decision

No policy changes on this evidence. The measurable next step is to widen the debt inventory beyond the four deterministic notice categories, because a debt observable whose whole population is four notices cannot discriminate between a healthy repository and an unexamined one.

That step is taken in Five Committed Executables That Nothing Demonstrates Ever Ran, which adds a second deterministic category and five findings.